We present a new approach to structural identification of complex dynamic systems by means of a fuzzy modular network. With all its advantages, extraction of structurally explainable models have been the envy of most system identification applications. A method based on fuzzy dynamic modeling in a pre-selected feature space is used in the self-organizing decomposition/modularization of the problem domain. A case example in a two-link manipulator arm is presented to further illustrate the ideas.
展开▼